d-Matrix today announced d-Matrix Squadron, a fleet of multi-rack systems offering megawatts of highly efficient inference compute for running AI models with extreme speed and energy efficiency at data center scale. Benefiting from d-Matrix's unique memory-centric compute architecture, Squadron clusters are ideal for AI labs, enterprises, and sovereign organizations with latency-sensitive AI applications including coding, personal agents, cybersecurity, and high-frequency trading.

Developed in partnership with Arista, Broadcom, and Supermicro, Squadron has already been deployed by d-Matrix in its Arizona data center for internal research and development and is now shipping to priority customers for deployment in their own data centers. Additionally, Squadron's building blocks—d-Matrix Corsair servers and d-Matrix Corsair SquadRacks—are now generally available.

Back in the now distant 2018, the semiconductor industry had three major players in contract leading-edge node manufacturing: TSMC, Samsung, and GlobalFoundries. At that time, Intel did not accept external clients. However, as node design became increasingly expensive—requiring multi-billion dollar investments—and more complex, with many players encountering yield and manufacturing difficulties, GlobalFoundries announced that it was stopping its 7 nm node development indefinitely. This left only two companies in the leading-edge race, but GlobalFoundries continued to develop unique semiconductor solutions that did not focus on smaller transistor sizes.

Today, the company is back at a similar stopping point but with a unique solution called FDX Fusion, based on fully-depleted silicon-on-insulator (FD-SOI) technology. Unlike traditional CMOS, GlobalFoundries takes substrate isolation a step further by creating a fully insulating layer beneath the channel. This enables highly specialized capabilities, such as ultra-low-power in-memory computing, seamless integration of embedded non-volatile memory (like eMRAM), and highly efficient dense logic designs. With FDX Fusion, GlobalFoundries claims 7 nm-class node characteristics in terms of size and twice the transistor density of the first-generation 22FDX node, which was designated as a 22 nm-class design.

Marvell Technology, Inc. , a leader in data infrastructure semiconductor solutions, today announced the second-generation Marvell Structera X 3406 and Marvell Structera X 3506 CXL memory expanders, designed to help cloud, enterprise and AI infrastructure providers build more efficient, scalable and higher-performance systems for the next era of AI and computing. Designed to address the evolving needs of leading hyperscale datacenters, Structera X addresses the most demanding data center workload requirements.

AI and ML workloads are rewriting the rules of data center design. Growing LLM sizes, expanding context windows and growing KV-cache requirements are driving unprecedented memory demands that traditional server architectures simply cannot meet. Memory is a critical constraint of AI performance and infrastructure efficiency.

Astera Labs, Inc. (Nasdaq: ALAB), a leader in semiconductor-based connectivity solutions for rack-scale AI infrastructure, today announced Orion: a modular rack-scale reference design with production-grade connectivity that allows customers to choose their own GPU and CPU. To be offered through Astera Labs' ODM and OEM partners, the Orion production-grade reference designs are expected to help enable neocloud and enterprise customers to deploy AI infrastructure faster and run it more efficiently, turning racks built on open standards into token-efficient AI factories.

For neoclouds and enterprises to economically serve inference at scale, they must secure sufficient accelerator and CPU capacity while delivering predictable, low-latency performance across increasingly complex AI infrastructure. But because they must commit to that capacity long before they know which accelerator will be available or affordable for their workloads, tying early investment to a single platform carries unnecessary cost and risk. At the same time, ODMs and OEMs are forced to repeatedly solve complex integration, validation, and software challenges for every new accelerator platform—slowing deployments when inference capacity is in critical demand.

INNO3D announces the AGS-4UMGX-R1 NVIDIA MGX 4U GPU Server, a high-performance platform built for AI training, AI inference, large language models (LLMs), digital twins, and high-performance computing (HPC) workloads.

Powered by dual Intel Xeon 6 processors and supporting up to 32 DDR5 RDIMMs, the AGS-4UMGX-R1 delivers the scalability required for modern AI deployments. The system supports up to eight NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs and features the NVIDIA ConnectX -8 SuperNIC with eight 400 Gbps QSFP ports for ultra-fast AI networking.

Today, AMD is introducing the next AMD Ryzen AI Halo Developer Platform, powered by the AMD Ryzen AI Max+ PRO 495 processor.

With up to 192 GB of unified memory, support for 300-billion-parameter and larger models, and a curated software experience, the new platform gives developers and creators the freedom to build, experiment, and run more AI locally. Choose your models, frameworks, operating system, and workflows, then put them to work on your terms.

Activision has reportedly issued a DMCA takedown against a set of unofficial Call of Duty browser ports made by developer Luckey Faraday. Faraday announced it on X, saying the publisher went after him "for porting CoD games to the browser," that every related post was removed, and that every game is now offline. He also asked whether he is the first "vibe coder" to get a DMCA notice from a major gaming company. The notice itself hasn't been shared publicly, so the exact claims are not known, and Activision doesn't seem to have commented so far. The projects in question are Claude of Duty: Vibe Slops II, a Black Ops 2 multiplayer recreation, and Kino der Toten, a browser version of Black Ops Zombies. Both ran on Three.js and were built with the help of AI coding agents.

Vibe Slops II offered free-for-all matches against six bots on Hijacked and Nuketown 2025, while the Zombies project had Kino der Toten fully playable solo. The project pages state that the maps, models, and animations were exported from the original PC games, which probably didn't help Faraday's case, even with both labeled as non-commercial fan projects. The Kino der Toten repository was still accessible on GitHub at the time of writing, so it looks like the takedown hit the hosted games and the posts promoting them. It comes less than two weeks after Take-Two sent a cease-and-desist to the modder behind the GTA 5 Switch port, and shortly after a free browser port of Halo: Combat Evolved showed up.

NVIDIA has reportedly stopped allocating GB202 silicon to the GeForce RTX 5090, according to leaker MEGAsizeGPU (@Zed__Wang) on X. "No more GB202 for 5090, confirmed," the leaker posted, adding in a follow-up that "all future GB202 supplies are RTX Pro exclusive." GB202 is the largest "Blackwell" graphics chip, and the RTX 5090 doesn't even max it out. The same chip powers the 96 GB RTX PRO 6000 with 24,064 CUDA cores, as well as the recently unveiled RTX PRO 5500, which pairs the RTX 5090's 21,760 CUDA cores with 84 GB of memory. Apparently, NVIDIA would rather put every GB202 it gets into those far pricier workstation cards. The company hasn't confirmed any of this as of the time of writing.

A second leaker, hongxing2020, posted "5090 EOL, 5080 24GB ready" alongside an apparent factory notice saying NVIDIA will no longer sell any RTX 5090 model, China-market D variants included. The same notice says an RTX 5080 24 GB will become the top GeForce card and the 16 GB version will be discontinued. That sounds a lot like the RTX 5080 SUPER, which was rumored with 24 GB of GDDR7 on the same GB203 chip, although it isn't clear if the two are the same card. SUPER cards reportedly reached board partners in July, only to be put on hold over GDDR7 pricing. As for the RTX 5090, I doubt NVIDIA would formally call this an end-of-life (EOL) move, since the chips are simply being redirected to a more profitable product line.

Frore Systems today announced LiquidJet Diamond, a new AI Factory liquid cooling coldplate featuring an integrated diamond spreader. The accelerating demand for AI Tokens, coupled with energy scarcity, is spurring hyperscalers to find new ways to increase the efficiency of AI Factories. LiquidJet Diamond Coldplate boosts AI Tokens/Watt and revenue by 35% compared to traditional skived coldplates, with the increased revenue flowing directly to profit.

LiquidJet Diamond builds on the previously announced Frore Systems LiquidJet coldplate. LiquidJet coldplate employs a unique 3D ultra short-loop multi-stage design which can be precisely customized to the power map of each GPU, including extreme hotspots. This precision is critical as GPUs grow in complexity and power draw to meet AI demand. LiquidJet delivers 12°C reduction in GPU die temperature which directly translates to 25% higher tokens/watt efficiency. LiquidJet Diamond adds an additional spreader with integrated diamond wafers to the coldplate bottom, positioned precisely at the hotspot locations. LiquidJet Diamond delivers an additional 10 °C reduction in hotspot die temperature, boosting increase in Tokens/Watt to 35%.

iBUYPOWER, the leading system integrator in North America specializing in high-performance prebuilt and custom gaming PCs, announced today that it is expanding its iBUYPOWER PRO Workstation lineup with the reveal of an all-new mini-PC - the iBUYPOWER Pro AMD AI Halo Mini PC, which is built on AMD's Ryzen AI Max+ PRO 495 Series processors. Available for purchase in Q4 2026 starting at $3999, the iBUYPOWER Pro AMD AI Halo Mini PC enables next-gen workstations and desktops to deliver exceptional ROI with robust security features while maintaining high performance.

COMPACT FORM FACTOR BUILT WITH iBUYPOWER'S HIGH-QUALITY STANDARDS
The new iBUYPOWER Pro AMD AI Halo Mini PC easily fits into any user's existing setup, measuring 70x198x200 mm, making it perfect for discreet or minimalist spaces. For those who prefer mini-PCs as their main system, the iBUYPOWER Pro AMD AI Halo Mini PC uses AMD Ryzen's AI Max+ PRO 495 Processor. This lets users easily render complex AI models, enable instant ray tracing and interactive simulations alongside local AI inference, and seamlessly tackle complex data and engineering projects without using third-party cloud services.

Kioxia America, Inc. has introduced its first E1.L form factor SSD built with Kioxia BiCS FLASH generation 8 QLC flash memory. Designed for high-density 1U servers and read-intensive storage applications, the new Kioxia LD4 Series targets hyperscale data center-class workloads and scale-out environments.

As data requirements continue to grow, hyperscale data centers continue to increase the storage capacity of their highest density systems. The Kioxia LD4 Series brings high-capacity QLC flash to the E1.L form factor for read-intensive applications in these environments.

Senao Networks Inc. today announced the launch of its new x86 enterprise server portfolio, featured on the company's Senao computing website at www.senaocomputing.com. The portfolio includes the SR810, SR710, SR610, SE210, and SE110 and is designed to support a broad range of enterprise computing and AI workloads, including generative AI, AI inference, model fine-tuning, data analytics, visualization, and intelligent edge computing. NVIDIA RTX GPU support is available in selected configurations.

Bringing AI Closer to Enterprise Data
As enterprises accelerate AI adoption, organizations increasingly need computing infrastructure that can address different workload requirements while maintaining greater control over sensitive corporate data.

OnlyOffice presents the major autumn update 10.0 with context-aware AI tools directly in the editors, new modular Apps, and a significant security upgrade. Around 50 new features and more than 1,000 fixes for the online, desktop, and mobile editors.

AI that understands your documents, in every editor
OnlyOffice 10.0 puts context-aware AI inside the document, spreadsheet, presentation, and PDF editors. A dedicated AI tab with built-in chat lets teams draft, summarize, translate, clean and format without leaving the file or copying content into a separate tool.

As Artificial Intelligence (AI) evolves toward Super Intelligence (SI), intelligent computing is entering a new phase—one in which increasingly capable AI is moving closer to endpoints and edge environments. This shift is accelerating demand for local large-model processing, data privacy, low-latency performance, and greater control over AI workloads. In response, Minisforum is advancing its Local AI strategy from high-performance AI systems toward scalable local computing platforms. The new MS-S1 MAX-P495 is the latest step in this evolution, bringing powerful local AI computing to the edge and helping bridge today's AI era with the emerging era of Super Intelligence.

From 395 to P495: More Headroom for Local AI
Following the MS-S1 MAX powered by the AMD Ryzen AI Max+ 395, Minisforum is introducing the MS-S1 MAX-P495, powered by the AMD Ryzen AI Max+ PRO 495. Built on the Zen 5 architecture, the processor features 16 cores and 32 threads, with clock speeds of up to 5.2 GHz, and integrates Radeon 8065S Graphics. It delivers up to 55 TOPS of NPU performance and 131 TOPS of total AI compute, targeting AI enthusiasts, developers, geeks, and professional creators with higher demands for local high-performance computing.

Sony Interactive Entertainment has a new patent application out that describes using machine learning to change how NPCs and bosses behave, based on how the person holding the controller plays. Titled "AI-adapted non-player character for video game system" (US 2026/0295429 A1), it was filed on March 31, 2025, published by the USPTO on October 1, and spotted by Patentlyze. The system collects player data and game data from an encounter with an NPC, feeds both into a machine learning model, and uses whatever the model returns to modify that character.

The data Sony lists is quite granular. On the player's side, it covers reaction time to boss attacks, the success rate of dodges, blocks, and parries, the weapons and consumables used, and even the number of retries. On the boss's side, it tracks attack patterns and how often the player avoided or countered them. Combat is only part of it, since the filing also mentions dialogue, appearance, and NPC voices that shift in pitch and volume. Sony says this content could be generated before a game runs, in real time during play, or afterward, so a boss that adapts mid-fight is one of several scenarios covered.

MSI, a global leader in high-performance and AI computing, today announced that the Prestige N16 Flip AI+ is now available for pre-order through MSI-certified partners in select markets.

Prestige N16 Flip AI+ is powered by NVIDIA RTX Spark, uniting NVIDIA AI, RTX graphics, and unified memory in one powerful and efficient thin-and-light Windows PC. Designed to run larger AI models locally—without cloud dependency—and move between creative work, development, and immersive gaming without missing a beat. Connect apps, tools, and workflows into one seamless experience. Private by default. RTX Spark brings NVIDIA's full-stack AI platform, ready for the AI workloads of today and whatever comes next. Delivered on the familiar, trusted, and secure Windows platform.

Dell Technologies announces an expanded portfolio of Windows PCs and workstations built for the age of agentic AI. Together with NVIDIA and Microsoft, Dell delivers agentic AI across every stage of the creative and development journey, spanning a premium creator laptop, compact creator desktop and an AI workstation within the Windows environment.

NVIDIA RTX Spark comes to the XPS and Dell lineup for the first time, powering the new XPS 16 Creator Edition laptop and Dell Creator Edition Desktop. Paired with Windows, RTX Spark brings advanced AI, creative and gaming workloads together on a platform optimized for its performance. Dell Pro Precision with GB300 will be available on Windows, bringing always-on AI compute into the environment enterprise teams already depend on.

Samsung's upcoming HBM4E memory has reportedly passed testing for NVIDIA's next-generation "Rubin Ultra" AI GPU accelerator family. This is great news for the company, as its previous HBM3E didn't meet NVIDIA's requirements, leading NVIDIA to choose Samsung for GDDR7 only, instead of HBM3E. The new HBM generation brings significant advancements, and Samsung is one of the leading manufacturers with designs currently running at 14 Gbps per pin and 2,048 pins per HBM4E stack. Samsung can achieve a maximum bandwidth of 3.6 TB/s, and at 16 Gbps, this maximum memory bandwidth will increase to 4 TB/s for a single HBM4E stack.

With about a dozen memory stacks per AI accelerator or GPU, the total memory bandwidth will be remarkable. The company is currently offering this memory in a 12-layer (or 12-high) configuration with a memory density of 48 GB per stack. However, customer requirements will also include 8-layer 32 GB and 16-layer 64 GB stacks in the future.

AMD held a press briefing to share how its "Agentic PC" push is going, and it brought a fair number of figures along. Apparently, the company has shipped over half a million Agentic PCs across more than 50 designs, alongside tens of millions of AI PCs across more than 250 designs. Of course, the term "Agentic PC" is AMD's own. It covers any PC whose CPU, GPU, and NPU can run AI agents locally, and AMD says Ryzen AI Max+ systems with 64 GB of memory or more give the best experience. The Ryzen AI Max 400 series is pitched as the "next-gen Agentic PC" for 2026, and most of the briefing focused on its flagship, the Ryzen AI Max+ PRO 495 "Gorgon Halo" with 192 GB of memory.

Samsung Foundry may be on the verge of securing a significant deal with AI lab Anthropic to manufacture AI accelerators. According to the latest report, Anthropic is developing its custom AI accelerator on Samsung's SF2 2 nm node, which will initially be produced in a lower volume of around 200,000 chips. The design and implementation of this ASIC are being handled by Broadcom, targeting around 4,000 TOPS of INT8 compute, with a TDP of 1,300 W per chip. This will be supported by either 192 GB of HBM3E or even 288 GB of HBM3E, if the design allows. The current production volume is low, as ASIC projects are typically counted in millions of chips. However, Anthropic is likely evaluating the project's viability, which requires a complete custom infrastructure with networking, rack design, power supply, cooling, and many other factors.

The current Samsung SF2 node is built using Low-NA EUV lithography and is manufactured in South Korea, with a new facility soon to be established in Taylor, Texas. For packaging these chips and integrating them with HBM, Samsung is likely to employ its I-Cube technology, which rivals TSMC's CoWoS and allows for 2.5D packaging. Combined with the 2 nm dies, Samsung can position itself as a serious manufacturing plant for external customers and slightly close the gap with TSMC and its contract manufacturing business.

AMD will "substantially" increase its chip supply in 2027, CEO Lisa Su told reporters in Taipei on Tuesday, while admitting that demand is still running ahead of what the company can ship. Su is on her third trip to Taiwan this year, meeting partners such as Foxconn, Quanta, Acer, and ASUS, with TSMC also on the schedule. She is reportedly heading to South Korea later this week to sit down with memory suppliers Samsung and SK Hynix. Here's what she said:

We've been able to increase our supply as we've gone through 2026, and we're going to substantially increase our supply in 2027, but we can definitely use more. There's very, very high demand for the next several years. We can see very high demand, and so we need more advanced wafer capacity.

Following what appears to have been a series of unfortunate events that led to a roadside laptop find resulting in confirmation that Microsoft will be launching a 20-core NVIDIA RTX Spark Surface Laptop, we now appear to have confirmation that Microsoft will also be launching a Surface Laptop Ultra on the other end of the Spark spectrum. According to Windows Central, Microsoft will reveal the full RTX Spark Surface line-up on October 7 at a special event in San Francisco.

Alongside the 20-core variant, the Windows Central report claims that Microsoft will launch a Surface Laptop with NVIDIA's 18-core RTX Spark. It is also claimed that the 18-core SKU will start with 24 GB of unified memory, while the 24-core version will start with 32 GB of memory. There will reportedly be a performance demonstration that shows off how the new Arm SoC delivers the same level of performance on battery life as it does when plugged into the wall.

Dell has released a patch for its Dell System Update (DSU) software, version 2.3.0.0, to address a critical vulnerability affecting the tool. According to the latest DSU update information, several CVEs are present that impact the DSU command-line tool. The highest-rated CVE-2026-86360 involves a full system privilege escalation in DSU versions prior to the latest 2.3.0.0. This issue arises from an improper limitation of a pathname leading to a restricted directory, allowing potential attackers to execute code with root privileges. This vulnerability compromises the underlying operating system's security and becomes a significant concern.

DellDell System Update, versions prior to 2.3.0.0, contains an Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal') vulnerability. An unauthenticated attacker with remote access could potentially exploit this vulnerability, leading to Filesystem access for attacker. This vulnerability is considered critical because it can be leveraged by an unauthenticated attacker to execute arbitrary code with root privileges. Successful exploitation may allow complete compromise of the vulnerable application and underlying operating system. Dell recommends customers upgrade at the earliest opportunity.

The use of AI-generated art to accelerate video game development is becoming more and more common, with massive game studios like EA and Capcom openly acknowledging that they will be using AI to streamline their game development workflows, despite public outcry about quality and concern for the human cost in an industry already rife with layoffs.

Addressing a question in a recent interview with Nikkei regarding how game production will change in the face of AI, SEGA's senior executive officer and deputy chief business officer, Tsuyoshi Saito, said that the company is already using AI to improve efficiencies at the company, specifically when it comes to business idea generation and publishing processes, but that it will not be using AI for creative aspects of game development.

Qualcomm announced today that it is officially licensing technology from Chinese technology giant Huawei. In a deal estimated to be worth $6.9 billion upon completion, Qualcomm will license various technologies from Huawei to integrate across its product portfolio. According to official releases, these technologies include computing, AI, networking, and more. Bloomberg reports that the deal also covers Huawei's latest LogicFolding silicon design. This is a significant win for Huawei, whose R&D teams have spent years designing a way to bypass sanctions and overcome manufacturing technology limits. By stacking silicon vertically across two sheets, LogicFolding dramatically reduces the distance between compute elements, increases efficiency, and significantly improves transistor density. The result is a 55% increase in transistor density and a 41% boost in power efficiency.

Update, 06:30 UTC, October 6: Qualcomm issued a statement that dismantles rumors that the company is licensing LogicFolding technology.

QualcommQualcomm and Huawei entered into a multi-year, broad patent license agreement that includes cross-licenses across a range of technology fields. The terms are confidential, but reports characterizing Qualcomm as a net payor on the license agreement are incorrect. In addition, assertions that the agreement is related to LogicFold are not accurate. Separately, Qualcomm has agreed to purchase certain Huawei non-cellular U.S. patents across multiple technology areas.